Simulation

Institut Mines-Télécom
Outline

Introduction

Multiagent-based Simulation approach

Multiagent simulation platform

Turtlekit

Institut Mines-Télécom
 Introduction

Institut Mines-Télécom
 Simulation
Introduction
Simulation
• “The process of designing a model of a real system and conducting
experiments with this model for the purpose either of understanding
the behaviour of the system or of evaluating various strategies
(within the limits imposed by a criterion or a set of criteria) for the
operation of the system.” (Shannon 1976)

• “Simulation is the discipline of designing a model of an actual or

theoretical physical system, executing the model on a digital
computer, and analyzing the execution output” (Fishwick 1994)

Model
• “To an observer B, an object A is a model of an object A to the extent
that B can use A to answer questions that interest him about A. “
[Minsky, 1965]
R. E. Shannon. Simulation modeling and methodology. In Proc. of the 76 Bicentennial Conference on Winter Simulation, pages 9-15, 1976.
P. A. Fishwick. Computer simulation : Growth through extension. In Society for Computer Simulation, pages 3-20, 1994
M. L. Minsky. Matter, mind, and models. In Proc. of the Intern. Federation of Information Processing Congress, vol. 1, pages 45-49, 1965.

Institut Mines-Télécom
 Simulation
Introduction
Simulation supports
• Understanding, Exploration, Clarification
─ to understand the behavior of the reference system thanks to a model
that is considered as a miniature reproduction of the reference system.
• Validation, Assessment, Verification
─ to test an hypothesis of the reference system, to validate or to certify the
underlying theory.
• Control, action, control
─ to support a decision process or a control that will influence the state of
the real reference system.
• Forecast, Prediction, Anticipation
─ to predict the possible evolutions of the reference system following
evolutions or disturbances.
• Communication, Formation, Visualization
─ to show and share the model of the dynamic of the reference system.

Institut Mines-Télécom
 Multiagent-based Simulation
approach

Institut Mines-Télécom
 Simulation
Methodologies
Fishwick
• The model design associates the real system with a
representation of this system (the model).
─ This model is built from real observations (objective) or knowledge
(subjective).
─ Data are usually formalized using formal semantics or mathematical
logic to reduce ambiguities as much as possible.
─ It is then converted to algorithms,
• The model execution phase is the processing of the algorithm to
produce numerical outputs.
• The execution analysis phase, deals with the analysis and
confrontation of the results of the program with the behaviors
observed in the model.
Fishwick, P.: Computer simulation: growth through extension. IEEE Potential February/ March (1996) 24 to 27
Drogoul, A., Vanbergue, D., & Meurisse, T. (2003). Multi-agent based simulation: Where are the agents?. In Multi-agent-based simulation II
(pp. 1-15). Springer Berlin Heidelberg.

Institut Mines-Télécom
 Simulation
Methodologies
Gilbert and Troitzsch
• Refine the Fishwick
proposal with the addition
of the model building
phase
• The initial model is written
into a computer program:
the operational model
─ Adaptation to the simulator
─ There are both operational Drogoul, A., Vanbergue, D., & Meurisse, T. (2003). Multi-agent
based simulation: Where are the agents?. In Multi-agent-based
models and Simulators. simulation II (pp. 1-15). Springer Berlin Heidelberg.

─ The differences between

models introduces bias.
Gilbert, N., Troitzsch, K.G.: Simulation for the
Social Scientist. Open University Press (1999)

Institut Mines-Télécom
 Multiagent-based Simulation approach
bottom-up modeling approach
Microscopic level: simulation of the behavior of the
components of the real system.
• The components: the agents
• Their relation: interaction and organization at a micro level.
Macroscopic level: Observation, Analyze of properties of
the multiagent system.

Example: Ants
• Micro level: ants are agents which put
pheromones in the environment
• Macro level: the shortest path

Institut Mines-Télécom
 Multiagent-based Simulation approach
overview
A Multiagent-based Simulation (MABS) is a microscopic simulation
model
• A Multi-agent system: the multiagent model of an actual or theoretical physical
system
• Simulation: controls of the evolution of the model in time.

Advantages
• MABS supports
─ Multi-level modeling:
• Different models of “individuals”: from simple entities to more complex ones.
• Different levels of representation: “individuals” and “groups” within an unified conceptual
framework.
─ The simulation of complex systems:
• Structure preserving modeling of the simulated reality,
• simulation of proactive behaviors,
• Parallel computations,
• Dynamic simulation scenarios
Limits
• Computation costs

Institut Mines-Télécom
 Multiagent-based Simulation approach
Contributors
The Thematician (expert of the domain)
• Role: Defines the intention of the simulation process.
• Result: the domain model which describes the multiagent model of the reality.
The agents are informally associated to the components of the system and their
relations are identified (interaction, organization).

The modeler
• Role: He translates the knowledge of the thematician.
• Result: the design model where the agents are a refinement of the agents
in the domain model. Their properties are expressed using concepts taken
from multiagent domain (behavioral model, communications, …)

The Computer Scientist

• Role: He designs the operational model and writes the computer program.
• Result: the computational system where agents are computational agents.
Drogoul, A., Vanbergue, D., & Meurisse, T. (2003). Multi-agent based simulation: Where are the agents?. In Multi-agent-based
simulation II (pp. 1-15). Springer Berlin Heidelberg.
Institut Mines-Télécom
 Multiagent simulation platform

Institut Mines-Télécom
 Multiagent, simulation platforms
Typology
The operational, simulated model can be executed on a
• Generic multiagent platforms
─ Advantage: the computer scientist knows his environment, i.e. the platform
and the related multiagent model.
─ Limit: The platform must be adapted (or not) to support the simulation,
─ Example: JASON, JADE (Tapas, PlaSMA), MASH, MADKIT (Turtlekit)
• Generic simulation platforms
─ Advantage: the computer scientist can use the same environment for different
design models.
─ Limit: a new operational model has to be built for each new simulations.
─ Example: MASON, SWARM, GAMA, CORMAS, TURTLEKIT, REPAST,
NETLOGO, …
• specialized simulation platforms
─ Advantage: some parts of the operational model can be already available.
─ Limit: adaptation to a new platform.
─ Example (traffic simulation platform): Archisim ,MATSim, MITSIMlab, …

Institut Mines-Télécom
 Multiagent, simulation platforms
Components
Components of a multiagent platform

A scheduler
• A temporal model: discrete, continuous, event
• A scheduling policy
─ Synchronization of agent evolution
─ Simulation of the simultaneity

Institut Mines-Télécom
 Scheduler
Temporal models
Discrete time model
• Time advances in discrete step, which are integer multiples of some basic period
such as 1 second, 1 day or …
• If the state at time t is q and the input time t is x, then the state at time t+1 will be
δ(q,x) and the output y at time t will be λ(q,x)
─ δ is called the state transition function
─ λ is called the output function

Discrete Time Simulation

Ti = ti, Tf = tf
x(0) = v0, …, x(9) = v9
q(0) = q0
t=Ti
while (t <= Tf) {
y(t) = λ(q(t),x(t))
q(t+1) = δ(q(t),x(t))
t = t+1
}

Zeigler, B. P., Praehofer, H., & Kim, T. G. (2000). Theory of modeling and simulation: integrating discrete event and
continuous complex dynamic systems. Academic press.

Institut Mines-Télécom
 Scheduler
Temporal models
Discrete event models
• appropriate for those systems for which changes in system
state occur only at discrete points in time.
• A discrete points in time is called an event.
Discrete Event Simulation
1. Initialize the state variables
2. Initialize the ‘collection of pending events’
3. Initialize the simulation clock
4. while (there are pending events to be handled){
Remove the pending event (E) with the smallest timestamp (t)
Set simulation clock to that time t
Execute the event handler for event E
}

Institut Mines-Télécom
 Scheduler
Multiagent scheduler

scheduling algorithm based on scheduling algorithm based on

a continuous temporal model a discrete temporal model
T duration of the simulation T duration of the simulation
time = System.time(); time = 0;
T = time + T Agents= {agents of the simulation};
Agents= {agents of the simulation}; while (time < T){
For (a: agent) For (a: Agents){
activate(a) \\activate(a)
while (time < T) a.ContextComputation()
time = System.time(); a.DecisionProcess()
a.actionProcessing()
}
time++
}

Institut Mines-Télécom
 Scheduler
Simultaneity problem

Discrete simulation
• Let t be the simulation time value and ai(t,q(t)) the action of the
ith agent following the current state of the simulated system q(t)
• How to ensure that q(t) will be the same for ai and ai+1 since
ai(t,q(t)) modifies the current state

Dealing with Multi-Agent Coordination by Anticipation: Application to the Traffic Simulation at Junctions.
A Doniec, S Espié, R Mandiau, S Piechowiak - EUMAS, 2005

Institut Mines-Télécom
 Simultaneity problem
Solution
No solution
• The most current solution,
• The action of an agent should not change the world in an important way. The
micro coordination problems resulting of the scheduling process are not taken
into account.
• The consequences of this choice have to be taken into account

The scheduling policy

• The activation order of the agents is randomized
• If the number of agents and simulation steps are important then no agent should
be advantaged.
• If the simulation must be replayed, the random process has to be taken into
account by the simulation model.

A dedicated mechanism
• The agents are activated in the same simulation state and the antagonism
between their action is resolved by a decision process.
• Influence / reaction model: The agents do not directly act in the simulation but
emit influences that are validated by the decision process.
Ferber, J., & Müller, J. P. (1996, December). Influences and reaction: a model of situated multiagent systems. In
Proceedings of Second International Conference on Multi-Agent Systems (ICMAS-96) (pp. 72-79).

Institut Mines-Télécom
 Turtlekit

Institut Mines-Télécom
 Illustrative Example

Prey and Predators

• A multiagent model
─ Environment: a grid
─ Prey: reactive agents who avoid
the predators
─ Predator: communicative agents
who coordinate to catch the preys
─ When three predators are around a
prey, this last one die
• Simulation
─ A scheduling process
• Temporal model
• Activation process

Institut Mines-Télécom
 Turtlekit
overview
Plugin of the Madkit platform dedicated to the
simulation
• Supports the Madkit organizational model: the
AGR model

─ Interaction are regulated by the organizational model

• The communications are regulated following the
organizational model

• The perception can be implemented following the

organizational model

Institut Mines-Télécom
 Turtlekit
overview
Supports the simulation of heterogeneous multiagent model
• The superclass AbstractAgent contains the methods for the
─ Management of the life cycle:
• activate(); end(); launchAgent(...) ;killAgent(...)
─ Communication management:
• broadcastMessage(…); sendMessage(…) ; nextMessage();
isMessageBoxEmpty(); receiveMessage(Message m);
─ Organization management
• createGroup(); leaveGroup(); requestRole() ; getRoles(); isGroup(…);
getAgentsWithRole()
Example
public void setup(){
playRole("predator");
ACLMessage m = new ACLMessage("INFORM",“I’m a new predator");
broadcastMessage("Turtlekit","HUNT","predator",m);
}

Institut Mines-Télécom
 Turtlekit
overview
Class Agent
• Inherits of the superclass AbstractAgent
• Implements the Runnable Interface
─ Methods “to control” the its thread
• exitImmediatlyOnKill() ; live() ; pause(int t) ; run().
─ Additional methods for communications
• waitNextMessage() ; waitNextMessage(long timeout)

Example
public void live() {
while (true) {
Message m = waitNextMessage();
if (m instanceof ACLMessage)
handleMessage((ACLMessage)m);
} }

Institut Mines-Télécom
 Turtlekit
overview
Class Turtle
• Inherits of the superclass AbstractAgent
• Do not implements the Runnable interface
• Additional methods
─ Related to the simulation process
• setup, activate, end
─ A turtle is a situated agent, he has methods to
• To be located in the environment
o setX, xcor, dx, distance, towards, getHeading, private boolean catched(){
• To move in the environment int cpt=0;
o moveto, fd, home, turnLeft, turnRight for(int i=-1;i<=1;i++)
for(int j=-1;j<=1;j++)
• To perceive the environment
if (! (i==0 && j==0) ){
o countTurtlesAt, countTurtleHere, turtlesAt, turtlesHere
Turtle[] tur = turtlesAt(i,j);
Example if (tur!= null && tur.length>0 &&
public void setup(){ tur[0].isPlayingRole("predator"))
playRole("predator"); cpt++;
randomHeading(); }
setColor(Color.red); if (cpt>3) return true;
return false;
if (countTurtlesHere()>0)
}
fd(1);
}

Institut Mines-Télécom
 Turtlekit
Simulation
Pseudo activation algorithm
time = 0; Turtle = {turtles of the simulation}; T duration of the simulation
While (time < T)
For (t: Turtle)
currentAction = scheduler.getCurrentActionTurtle(t)
nextAction = activate(t,currentAction)
scheduler.setCurrentActionTurtle(t,nextAction)
time++

Institut Mines-Télécom
 Madkit
Message

Message

StringMessage

KernelMessage

madkit.kernel

ActMessage ControlMessage ObjectMessage

ObjectMessage XMLMessage
ACLMessage KQMLMessage

madkit.messages
Institut Mines-Télécom
Madkit
Message

Message
• getCreationDate; getReceiver; getSender
ActMessage
• getAction; getContent; getFieldValue;
getInReplyTo; getObject; setContent; setField;
setInReplyTo; setObject
ACLMessage
• getAct, getPerformative; setPerformative;
getReceivers;removeReceiver;clearAllReceiver.

Institut Mines-Télécom